Episode Transcript
[00:00:00] Speaker A: Hello. Welcome to another episode of Cracking Copy. I'm Ella Hoyos of Flurry Marketing and my Cracking Copy co host is Minnie, but I'm not with her today. I'm with John Gillam from Originality AI and this is part of our series that we are doing on copywriting and AI. And I've invited John on because he has founded his own company which is all in terms of AI detection and plagiarism. He's going to tell us a bit more about the tool that he has created and what it's for and what we need to know. So welcome, John. Welcome to Cracking Copy.
[00:00:33] Speaker B: Yeah, thanks, Ella, thanks for having me. Always, always happy to talk AI and AI and content marketing for sure.
[00:00:39] Speaker A: Yeah, well, me too, because I find it a hugely exciting new field, obviously very fast growing and rapidly changing and I'm just, I feel like I'm like a sponge at the moment. I just want to soak up everything there is to know about it or not everything, because, you know, it's moving way faster than I could. But I'm really keen to learn more about the potential of it and I see a huge opportunity there for all of us. I'm. I'm integr it into my daily practices. I'm also mindful of where this thing is going and what shape this is going to look like into the future. So there are risks, I'm sure. But first of all, tell me a little bit about yourself and how you're in this world.
[00:01:17] Speaker B: Yeah. So my background was a mechanical engineer at a school and then got into the world of sort of content marketing by building websites and looking to get traffic from Google. That grew into building up a content marketing agency. We're doing copy for, for other customers and then sold that and had seen through that wave, through that sale that sort of. There was this wave of generative AI coming. We were some of the earliest users of Jasper and then transparently selling that content as AI generated content. But there wasn't a great sort of way of you could put a policy in place around AI, but there wasn't a great way to sort of enforce that policy. And that was what brought us to the idea around originality of building a modern day QA QC tool. You know, cop plagiarism. Checkers have been around forever, but they hadn't evolved significantly for, especially for the needs of a volume publisher. And so that was kind of some of the starting, starting idea. And we ended up launching originality the weekend before ChatGPT launched, which was great timing. But also kind of. Kind of crazy.
[00:02:18] Speaker A: Yeah. Was that serendipity? Did that just happen by magic or had you known that ChatGPT was going to launch?
[00:02:24] Speaker B: No, no, def. Definitely by. By luck. And, and certainly if we'd known, we would have pushed the pace because that this was a sort of a side project at the time and would have pushed the pace a lot harder and a lot more aggressively had we. Had we known that the world was going to go for AI crazy.
[00:02:40] Speaker A: Yeah. Yeah. And within five days, hadn't the amount of AI ChatGPT users surpassed? I don't know. I think there was a trajectory, isn't it how long it took Facebook to reach the mainstream and how long it took all these other tools, and yet Chat GPT did it in five days.
[00:02:55] Speaker B: Yeah, exactly. Fastest growing app of all time.
[00:02:57] Speaker A: Yeah. Yeah. Well, I think it sounds like you launched at the perfect time, actually to me. Do you just refer to QA QC before I'm taking.
That means quality control?
[00:03:07] Speaker B: Yeah, it's sort of, yeah, quality assurance, quality control. Sort of anything that a copy editor does when they receive text from someone, we aim to build a solution that lets them do that process way easier. AI detection has been sort of the big thing in the last few years. But a detection, plagiarism checking, fact checking, readability, grammar, spelling, rhetorical guideline compliance, that's upcoming. But sort of anything that a copy editor does when they receive text from a writer, we aim to sort of provide tooling, some of it AI powered, to help help them do their job more efficiently.
[00:03:39] Speaker A: Yeah. And I guess it's all about the integrity of getting really good quality articles into a publication, whether that's an online publication or magazine, just to make sure that the quality is up there and meeting the standards of that particular publication. Okay, just tell me a little bit about originality. AI, how does it work?
[00:03:57] Speaker B: Yeah. So the AI detection, sort of the simple explanation around a detection is that it's like imagine, imagine there's an editor that sits in a room for a thousand years and reads millions of copies of human text and millions of copies of AI text, and they've got a, you know, through the roof IQ and are able to sort of like start to identify patterns that are hard for humans to recognize. But, but they are able to recognize patterns that human writers have, that AI has and ones that they don't have, and then is able to. We have a sort of another test to say how, how accurate is this tool now at figuring out kind of how, how likely that piece of content was written by AI or however it likely was written by a human.
[00:04:42] Speaker A: Is that, when you refer to that, is that your confidence score? You give things a confidence score. How confident it is that this is 90% AI and you know that sort of thing.
[00:04:52] Speaker B: Yeah, exactly. So it'll. The tool. The tool will now say, I think that was AI and I'm 90% confident. All right. I think that was human and I'm 70% confident.
That score is a confidence score. Doesn't necessarily mean that we think 70% of this content was AI generated and 30% was human generated. That we think that the model is saying, I think that was AI and I'm 70% confident.
[00:05:15] Speaker A: Okay. Okay. So, yeah, just. That's good for accuracy, so that you can get a gauge of how accurate it thinks it is. It allows for a level of uncertainty, I suppose, in the result.
[00:05:27] Speaker B: Exactly like we wish. We wish this was like, when we built this, we were hoping it would be very much like we're all used to plagiarism, checking where it's like, hey, here's the 10 words that were clearly copied from over here into here. I can see the link. I can see that that was published earlier. We would have loved to be able to produce something that created that level of clarity. What we have realized that, you know, classifiers, which is what this sort of form of AI is that we. That we built, aren't. Aren't highly accurate, but not perfect. So if you need sort of like an absolute certainty on this, like, was this article AI generated or not? It's like, it can give you confidence. It's so, you know, we were talking about the weather earlier, but similar to the weather where it's like, probably gonna rain, doesn't necessarily mean that it for sure is going to rain.
[00:06:08] Speaker A: Yeah, I get it. Okay. But what I'm really curious about is why do you. What do you think is the biggest problem for companies or individuals using AI generated content? Why is that an issue?
[00:06:20] Speaker B: Why do we. Why do we care? I think so. I think there's. There's a few reasons. There's the sort of simplest one is that there's a fairness component to it where if, you know, some. If, let's say a publisher or somebody hiring a writer is happy to pay them a hundred dollars, thousand dollars, you know, whatever that rate for that piece of content is, they're not super happy to find out that that was copied and pasted out of chatgpt in 10 seconds. And so if the publisher, whoever's hiring that writer is wanting is accepting AI content, then they should also be sort of capturing the efficiency or some of that efficiency from that decision. So that's sort of the fairness piece to it. And then there's the, the, the risk component. So publishing with AI content is riskier one, it's riskier because of the sort of factual accuracy. And it can be extremely confident at hallucinations which, which aren't as big a deal as they used to be, or people are getting better at dealing with them. But it'll certainly just make stuff up at times that is wrong. And then trying to sort of solve for that is, is a risk to the company that if AI is being used, there's a higher chance of factual inaccuracies making their way in. And then the biggest risk for the majority of our customers is in the eyes of Google. So Google has made it very clear in their actions that they are against the mass publishing of AI spam. And I think what's sort of interesting right now is that, you know, not all AI content is spam, but in 2025, anybody that's spamming content is definitely using AI to do so.
And that creates a risk of anyone else that is publishing AI content to get wrapped up in updates from, from Google. So the, so what, what's the problem is the fairness component. And then there's a whole group of other items that I just bucket under the risk component. So if you're publishing AI generated content, it is riskier than publishing human written content. And that decision on whether or not to publish riskier content should be made by the sort of, the decision maker of that flow of content. So whether that's the purchaser of the publisher, you know, whoever's in charge. But that risk, that decision to accept that risk should be, shouldn't be made at the, at the writer level, but more, more at the sort of business owner level.
[00:08:33] Speaker A: Okay, yeah, that's interesting. That's really interesting. I thought the job of an editor though is to actually do the fact checking. Or if you're a public publishing house, you may have a legal team and the copy will go through legal to pass some checks. So, you know, humans are flawed too. We don't always get it right. We might not, with the best intention, might not intend to make a mistake. But maybe we've written something that was poorly researched or we relied on doing a Google search to find our answers and we trusted the Internet and we've still turned in something that is perhaps, well, it needs a fact check. It's flawed in some way. So you know, humans can make those, those mistakes too. So surely there's a, there is a risk with that as well.
[00:09:12] Speaker B: Yeah, no, I think, I think for sure, for sure some of those, some of those risks exist. I think the, the risk of having somebody dial up the content publishing velocity because AI is getting used. So we've had, we've seen companies where they had a junior marketer that was pretty tech savvy that they didn't have any controls in place. And that junior marketer sort of spun up some, some technology to sort of mass publish blog posts and they totally killed their business because they lost and Google viewed that as spam and, and totally shut that shut the business down. So I think that's, that's an extreme example and I think you're, you're exactly correct that humans can make that, those mistakes as well. But as you sort of welcome in the efficiency of AI, it accelerates your exposure to that risk of some of those things that we talked about. The factual, if you're publishing velocity increases, then your rate of potential factual errors increases. And so I think that it's a bit of a spectrum of sort of like from, from low risk to high risk. And if you sort of just open AI with no controls in place and just say have at it and publish away, then, then that's a recipe for big problems for, for your site.
[00:10:24] Speaker A: And is this mainly, you know, any, are we talking about anybody who publishes third party content on their websites? It doesn't have to be an agency, does it, that would use a tool like originality?
[00:10:34] Speaker B: Yeah, I think, I think that's exactly it. Is anybody that anybody that publishes content that, that they got from somebody else, if they wrote it themselves, then that's one thing. But anybody that's functioning as a copy editor is exposed to this risk and that originality helps helps address. And so that could be that, you know, that could be the owner of a, of a pool store that has, has their sort of child writing content for them for marketing copy for them. And so they're, they're, you know, they don't think of themselves as, as an editor, but that's what they are performing the dude, the job of an editor. When they receive that text and put it up on their, on their site or they just let that third part, that third party kind of turn loose. So I think there's a lot of people that are hiring marketing companies unaware that a lot of the content that they are now getting is, is AI generated and that's a risk to them. That they're not currently, currently probably correctly sort of pricing into to their, their workflow.
[00:11:29] Speaker A: Okay, so presumably your advice is going to be run your content through a tool like yours, give you that confidence that it's, it's good to publish.
[00:11:38] Speaker B: So I think the first step is like some, some people will be okay with that risk. And so I think the, the first step for most, for anybody in this world right now is to understand where the use of AI is acceptable and not acceptable and then put in then controls to manage that originality is absolutely one of those, those controls to know if content was AI generated or not. But I think it's more important than that. The use of the tool is the sort of clarity and communication of the decision around where AI is allowed and not allowed in marketing copy.
[00:12:11] Speaker A: Yeah. Okay. So get clear on where you're comfortable, what your comfort level is with the use of AI. So one way that I've done this in my business is to create an AI policy. Now because I am using these tools, like many of us, to assist with things like social media, content creation, things like that, to speed up some of my processes for my business, I have created a policy which shows, which just talks about how I use it in my business and I've published that on my website now so that clients can see that I'm. Well, I'd like to be transparent about it and let's have a, let's have the conversation. You know, if, if a client isn't comfortable, if particularly I used to write for magazines and newspapers, I'm sure they don't want my AI content, you know, fair play to them, that's fine. Well, I'm perfectly capable of writing my own copy and content as well. So yeah, I think just being transparent and knowing not just from your perspective as a business what you're comfortable with, but what your clients or your employers are comfortable with and making sure that you communicate that I think is wise, isn't it?
[00:13:12] Speaker B: Yeah, no, that's, that's, I think that's exactly, exactly right. And, and yeah, having an AI policy and communicating it transparently is certainly, certainly key.
[00:13:20] Speaker A: Yeah.
Okay. Now it's interesting as well that Google have said this of anti, you know, against mass publishing of AI because they've got their own tool now, Gemini, and when you do a Google search inquiry, the first result comes above the paid results and the organic search results. Now you get Gemini's summary, you know, of what, you know, what it thinks you're looking for. That's pure AI and presumably that's all subject to all the same hallucinations and bias and other things that could creep into it. So you know, it's a bit from what you're saying, sounds a bit rich coming from Google. Like you're not allowed to do this, but we're gonna, we're doing it.
[00:14:00] Speaker B: Yeah. No, Google's facing, it's an interesting moment for them where, where they're sort of facing their Kodak moment, their sort of existential threat around technology that they sort of were fundamental in sort of bringing to life, which was sort of transformer technology and LLMs and will that potentially replace them. And so they have to sort of walk this line between being this very sort of like technology forward, AI forward company. But then if they kill the golden goose that is their search results, if their search results are filled with nothing but AI content, then why would people go to Google? Why wouldn't they just go to an ll, a different LLM? They sort of lose all advantage and also like longer term could, if the results were filled with nothing but AI, kind of kill that ecosystem that has created the data to train these LLMs on in the first place. And so it's this really sort of like multidimensional problem that they're faced with and the way that they seem to be sort of addressing it is being this sort of like very AI forward company on like the company side, on the search side being very AI cautious and sort of talking about AI spam and any mass produced content is viewed poorly and then taking some really significant actions against AI content in their search results and AI sites and then using AI to try and keep more of the clicks for themselves and sort of fight off the open AIs that are potentially going to disrupt them. So it's a challenging, challenging situation that they face that they're in.
And you know, Google doesn't want to lose. But I think the net result is that it's also hurting, hurting with Google capturing more and more of the clicks themselves.
[00:15:42] Speaker A: Yeah, and particularly people who are paid advertisers, people who use Google Ads. Their stuff's been trumped now by Gemini's search results and things like that. So yeah, it's very interesting and I think just watch this space. I thought it was odd when the Gemini results came in. It's not been that long, I don't think, but it's definitely the way the thing is moving.
Will, as a business owner, if you have a website and publish a blog or if you're a content marketing agency and then take on other Content from third parties like we talked about. Should you be worried if you're a, if you're a blogger, should you be worried about publishing AI generated blogs on your site? Should you be worried about Google coming to shut you down or something?
[00:16:25] Speaker B: So yeah, so I think my, my opinion that's backed by what we've seen Google have, have done to other sites is that yes, that Ma. Ma. I think it's, there's no question that mass publishing AI content is, is viewed as spam and Google is aggressively punishing sites that have mass published AI content where that line. And then I think it is safe to publish really high quality human generated content. And again sort of if we view this as a spectrum that like mass publishing, mass publishing AI content bad going to get you in trouble. Publishing really high quality human created content good. You'll, you know, you'll, you'll eventually get traffic and that, that'll hopefully work out. But then there's a, there's a whole line between that. What about, what about mass published human content that's low quality also? Probably not, not great. What about human content that AI has been involved in the creation of? Okay, like so I think it's, there's, it's, it's a sort of risk decision on what people are willing to accept and not every publisher has the same risk profile. And so you know, for, for ourselves at originality, we get a lot of traffic from Google. Now I have a content marketing team that's running really hard at it sort of. We're, we're over a million, a million visitors a month from Google. We have strict no AI policy with the exception of some of our technical team. Some of our AI research team are English as second language individuals. When they share some of their research findings they can use AI to help them with the content and then it all goes through an editor. And so that's sort of our nuanced approach to where it is acceptable and isn't acceptable. But to answer your question directly, publishing AI content is a riskier proposition for business owners than, than not publishing AI content.
[00:18:14] Speaker A: And what do we, what are you defining here as AI content? Because there's obviously with, I'm largely talking about generative AI where it can create stories for you, it can create lots of text for you, long form copy. But there are then the other AI tools that have been around a long time, like the grammarlys of the world that do the spell checking and the grammatical correction. Is your AI policy ruling that stuff out as well or is it just the generative AI content.
[00:18:40] Speaker B: So, so those tools do use generative AI as well. So if depend. It depends on the writers. Like you're, you're a writer. If your content goes through Grammarly, it should still get called human content because that was, that was probably not many changes were. Were needed. I'm a terrible writer when my content goes through Grammarly. It's like all ton of suggestions and then I just need to like it. It needs to do a lot of work to, to make that content clear. And so the, the sort of your content into Grammarly and out of Grammarly probably still called human. My content into Grammarly, a lot of, A lot of it needs to get rewritten by. By sort of Grammarly's LLMs and, and the outcome is a AI generated piece because it's how close to what I originally put in my thoughts. I might still use it, but it has definitely been heavily changed as a result and no longer recognizable to the initial initial content that went in because of the amount of changes that, that Grammarly suggestions needed to make to make my crappy content reasonable.
[00:19:42] Speaker A: Okay. Okay. And presumably you don't get penalized too much if you're just a bad typist as well. You know, I'm okay at writing, but no.
[00:19:51] Speaker B: So that's fine.
That's just correcting words.
[00:19:54] Speaker A: Yeah. Okay.
[00:19:56] Speaker B: You're not, you know, it's not totally rephrasing entire paragraphs.
[00:19:59] Speaker A: Yeah. It's not changing sentence structures and paragraph structures, that sort of thing. Okay. How should businesses and copywriters and you know, general writers use AI tools responsibly in a way that's going to benefit them and not penalize them?
[00:20:15] Speaker B: I think you, you said the key word earlier, but transparently. And so if you have a writer that is, that is transparent about the way that they do their work process and they use AI for the, the research.
They pull in some unique data from, from this location and they go and write the content. And that works for you. Great. You have somebody else that's a subject matter expert, but they don't love writing and they say I'm going to like put my ideas down and I'm going to have, have AI turn it into a consumable piece of content and that's okay with you then that, that's great. And so I think how it's flexible to every situation, but transparently between the writer and the, the end. The end user should be, should be. Should be understood.
[00:20:59] Speaker A: Okay. We talked about some of the risks, particularly reputational risks to companies are There any other risks that you see?
[00:21:07] Speaker B: So a risk for the writer. When we, when we launched, we'd worked with a ton of writers as part of the agency that we had sold and we really viewed sort of originality as a tool, tool to help defend the livelihood of writers. If there was, we sort of felt that if there was no tool to differentiate between AI content, human content, then a small and shrinking percent of the population of the world can write better than an AI and we, we felt like that that livelihood was going to be at risk if you were no longer able to tell the difference between human content and AI content. So I think that's a risk to writers of detection. If, if, yeah, if detection is not trusted or used then, then the sort of ability of how many, who else you're competing with to, for any writing assignment massively increases to anyone with a, with a ChatGPT login.
[00:22:00] Speaker A: Yeah, I love that because you're protecting the craft that is writing and it is a craft. It's an art and it's a science, depending on which end of the spectrum you're at. But there is definitely a lot of craftsmanship that goes into it. And we've seen in other industries how crafts are dying out because they've been replaced by machines like, you know, mechanics and even simple things like I like to sew, but I don't do needle and thread. I do the sewing machine, but we're losing our dexterity in our fingers. We don't learn that stuff at school as much anymore. So I think that's really important and it's great that you have recognized that. And just finally to end then, given tools like yours, do you believe that human writing is more important than ever?
[00:22:42] Speaker B: Yeah, more, more important than ever. I'm not sure. Do I think it's better than more powerful than AI? Yes, I think. Do you think it's less risky than AI?
I, I think there is a large, a large percent of copy editors that are able to be replaced by AI now. I, I, and is that, and some, some business owners are choosing to up that risk of AI content. And so I think it's a, Are they more valuable now than ever? I'm, I'm not, I'm not sure if that could be, can be true or not. I think AI is going to disrupt this profession and has and will continue to significantly. I think there is a large demand for good quality copy writers that AI has not, has not disrupted. And I think their, their value is as, I don't know if I would say More important than ever, but as important as, as ever. Which is, which is extremely, extremely important.
But I think there was certainly a portion of this sort of profession that is, is, is certainly going to be going to be challenged with, with generative AI coming.
[00:23:52] Speaker A: Yeah, absolutely. And a lot of the pressure comes from the need that, the perceived need to create more content, to post more, to be more visible, to be out there more. And I think that pressure, that does put pressure on writers to, to use tools like AI tools to help them do stuff faster. But yes, there's a very big argument that. Are you just adding to the noise there? What is it that's differentiating you? And if you're not really thinking about the words you're putting on the page and the writing and the messages you're trying to convey, then they're not going to resonate with the reader anyway. It's just going to be tuned out. There's more noise out there on the Internet. So I think there is real value for skilled writers and keeping your human, humanness in your writing for sure. Brilliant. Just a little bit more then about originality. Presumably it's a paid product. Where can we find out more about that?
[00:24:45] Speaker B: Yeah, so originality AI, there's a free tool for each of the sort of tools that are in the paid app. So AI detection, plagiarism checking, fact checking, readability, grammar, spelling, SEO optimization, so sort of any function that somebody does when they get text from, from a writer that you need to do to make sure that that article meets your standards and we provide tooling to, to help people do that and sort of make sure that they're, they're publishing content quality of really high order that will perform, perform well for them.
[00:25:18] Speaker A: Yeah, not just in Google search results, but reputationally it will perform well and just keep the standard of your work high. Yeah, I did have a play with originality before I came on and you can copy paste your text in just to test it out, can't you? I copy pasted some stories that I'd written and I got the green light on the ones I'd done myself and then it was very obvious to the tool which ones were edited or changed in using AI tools and they came up red and amber and orange. So I think, you know, as far as I was concerned, this is absolutely spot on. It knew I haven't tested it for the plagiarism, but I can see that if you use the paid version, upgrade to the paid version, you get that confidence test result and test score and there's a plagiarism check there as well, isn't there? So that's correct. Put the details in the show notes for anyone who's interested and then anybody else who might like to reach out to you. Where, where do you hang out online? Where can people find you?
[00:26:14] Speaker B: Yeah, LinkedIn a little bit or email jonriginality AI. I certainly live in live in email and Slack more than, more than social media. But LinkedIn would be the most most common place for me to be.
[00:26:29] Speaker A: Great. Okay. Actually that just brings me one final question I've got before we, before we end here in terms of Google. So Google have lots of products for the end user. I'm a user of Google Workspace. I use Gmail and lots of other people use Gmail as well. I'm just wondering much this AI content penalization, if you like would filter into emails. Do you know anything about that? Because I think recently there's been some changes to the way email filters work. The way that Google will identify whether something is spam content and whether it'll put it in your promotions folder or your primary folder. Do you think that AI content generated content is going to be penalized in email inboxes in the same way as it is online?
[00:27:13] Speaker B: Do you know anything really interesting question that I haven't, I haven't thought about. I don't know. We'll probably go and test that though because we run a lot of tests and we find it fascinating sort of to see how it, how it plays out. So yeah, I think we've seen for all the other questions we had a lot of data and time to think about it. I'm certainly guessing at the answer on this one. I think if AI gets used heavily for cold outreach emails and then content that, and then it's an AI classifier that's being used to identify what is and isn't spam or promotion that long, long form AI content structured in a promotional way will be more commonly tagged as more would be more common to be a spam email.
[00:27:57] Speaker A: I just think a lot of spam.
[00:27:58] Speaker B: Emails are so short that, that you need a minimum, you need a minimum content length to be able to. For detectors to work effectively. And so many emails are sort of like two sentences. So many spam emails are two sentences that it might be, it might be tricky. So I don't know.
[00:28:14] Speaker A: Yeah, I'm curious about that too because I know, you know, a lot of us, especially business owners, email marketing is a core function of what we do. We're encouraged to get our custom followers off social media and into our inboxes. And then when we've got an email list and we're emailing regularly, it's very tempting to lean into again into AI tools to generate your weekly email newsletter or whatever else. But if Google is playing hardball on this sort of stuff, then presumably it might be starting to filter out AI generated emails in the same way and sticking them more into the promotion tab and the thing, possibly. So, yeah, I'd be interested.
[00:28:49] Speaker B: Yeah, we'll test that. I don't know. I wouldn't say that I'm concerned enough at this stage that I would significantly change the way that if I was using AI for emails, would I. Would I be highly worried? I don't. I don't know.
[00:29:01] Speaker A: Okay. All right, well, if you do run any tests and do come back, I'd love to talk about it.
Okay. John, it's been really great having you on. Thank you. You've opened my eyes to these tools out there because it's not something I've been in the habit of using, but it certainly made me a lot more mindful about generating AI content. So thank you.
Appreciate that.
[00:29:22] Speaker B: All right, thanks, Ella. Thanks for having me.
